Direct current charging pile power device transient fault rapid detection method and system

Abstract

The invention discloses a direct current charging pile power device transient fault rapid detection method and system. The method comprises the steps of collecting a first feature of a direct current charging pile; processing and analyzing the collected first feature through a charging pile fault recognition model, detecting the change of each charging stage, and analyzing the change rule of the change; on the basis of the change rule, abnormal fluctuation of charging current and charging voltage in the charging process is monitored, and fault types are classified; and analyzing the abnormal fluctuation through an abnormal detection model, and determining a fault type in the charging process. According to the invention, possible power device faults or abnormal states can be rapidly detected, so that faults in the charging process can be intelligently sensed.

Description

Technical Field

[0001] The present invention relates to the field of power detection technology, and in particular to a method and system for quickly detecting transient faults of power devices in a DC charging pile. Background Art

[0002] Foreign countries have already conducted in-depth research in the field of charging equipment detection technology. Many companies have devoted themselves to the research and development of charging pile detection products and have achieved remarkable results. For example, Fluke, one of the leaders in the field of measurement, has successfully developed the 6658A AC and DC charging pile detection device. The device integrates a multimeter and a variety of test instruments, and has written an automated detection program. It can measure a maximum voltage of 1000V and a maximum current of 250A with a measurement accuracy of 0.05. The device is suitable for certification and testing institutions such as power companies and metrology institutes, and can efficiently complete the electrical performance testing of charging piles.

[0003] When optimizing the warning rules, the primary challenge is to comprehensively consider the changes in battery status in multiple scenarios. The diversity of charging scenarios causes the changes in battery charging SOC to be affected by many factors, such as charging rate, temperature, current, etc. The influencing factors in different scenarios may be completely different. Through large-scale data collection and analysis, a prediction model for charging scenario classification should be established using machine learning methods to more accurately predict potential charging safety hazards.

[0004] Another challenge is to solve the problem of factor identification for a large amount of unlabeled fault information. The lack of clearly recognized fault factors makes it impossible to classify or label fault information with certainty. It is necessary to apply feature engineering, cluster analysis and expert knowledge, use supervised learning and semi-supervised learning methods to build models, identify and classify fault factors, form a charging facility fault feature library, analyze and mine potential fault factors based on classification trees and other methods, and realize the mining of potential characteristic factors of charging pile faults.

[0005] The accuracy of predicting charging safety hazards is also an important challenge. The change in battery charging SOC is closely related to charging safety hazards, but the dynamic and complex nature of the charging process poses challenges to accurate prediction. It is necessary to adopt real-time data collection technology, combined with machine learning algorithms and model predictions, to establish a dynamic prediction model to classify and predict abnormal behaviors during the charging process in order to improve recognition and early warning capabilities. Summary of the invention

[0006] The purpose of this section is to summarize some aspects of embodiments of the present invention and briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section and the specification abstract and the invention title of this application to avoid blurring the purpose of this section, the specification abstract and the invention title, and such simplifications or omissions cannot be used to limit the scope of the present invention.

[0007] In view of the above existing problems, the present invention is proposed.

[0008] Therefore, the present invention provides a method and system for rapid detection of transient faults of power devices in a DC charging pile to solve the problem of how to classify and predict abnormal behaviors during the charging process.

[0009] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0010] In a first aspect, the present invention provides a method for rapid detection of transient faults of power devices in a DC charging pile, comprising:

[0011] Collect the first feature of DC charging pile;

[0012] The collected first feature is processed and analyzed by the charging pile fault identification model to detect the changes in each charging stage and analyze the change rules;

[0013] Based on the changing rules, monitor the abnormal fluctuations of charging current and charging voltage during charging and classify the fault types;

[0014] The abnormal fluctuation is analyzed by an abnormal detection model to determine the type of fault in the charging process.

[0015] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0016] The first characteristics include the charging current and charging voltage of the DC charging pile, and the charge state information of the charging vehicle.

[0017] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0018] The charging pile fault identification model comprises the following steps:

[0019] Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting were used as sub-classifiers;

[0020] Each sub-classifier calculates the probability of each type of fault occurring for the first feature of the input;

[0021] By taking a weighted sum of the probabilities of each type of failure occurring in the first feature calculated by each classifier, the failure type with the highest summation result is determined as the discrimination result of the recognition model.

[0022] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0023] The charging stage includes a start-up stage, a constant current charging stage and a charging completion stage;

[0024] The classification of fault types includes DC bus fault, charging pole abnormality, converter DC side outlet abnormality, load side fault and power supply side fault.

[0025] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0026] The method of detecting the changes in each charging stage and analyzing the changing rules thereof comprises the following steps:

[0027] If the charging power fluctuates abnormally during the startup phase, it indicates a DC bus fault;

[0028] If the charging power fluctuates abnormally during the startup phase, it indicates a power supply side fault;

[0029] If the charging power growth stagnates during the startup phase, it indicates a load-side fault;

[0030] If the charging power drops rapidly and greatly during the constant current charging phase and is lower than the preset safety threshold, it indicates that there is an abnormality at the DC side outlet of the converter;

[0031] If the charging power fluctuates frequently and cannot drop to the expected minimum point during the charging completion stage, it indicates that there is an abnormality between the charging poles.

[0032] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0033] The anomaly detection model includes using an unsupervised learning algorithm to model and train the current and voltage data.

[0034] As a preferred solution of the method for rapid detection of transient faults of power devices of a DC charging pile according to the present invention, wherein:

[0035] The analyzing the abnormal fluctuation by using an abnormality detection model includes inputting the real-time collected current and voltage data into the constructed abnormality detection model, identifying data points that do not conform to the normal pattern, and marking them as abnormal.

[0036] In a second aspect, the present invention provides a system for rapid detection of transient faults of power devices of a DC charging pile, comprising:

[0037] A collection module, used for collecting the first feature of the DC charging pile;

[0038] A detection module, used to process and analyze the collected first feature through a charging pile fault identification model, detect changes in each charging stage and analyze the change rules;

[0039] A classification module is used to monitor abnormal fluctuations of charging current and charging voltage during charging and classify fault types based on the change rules;

[0040] The analysis module is used to analyze the abnormal fluctuation through an abnormal detection model to determine the fault type in the charging process.

[0041] In a third aspect, the present invention provides a computing device, comprising:

[0042] Memory, used to store programs;

[0043] A processor is used to execute the computer executable instructions, which, when executed by the processor, implement the steps of the method for rapid detection of transient faults of power devices of a DC charging pile.

[0044] In a fourth aspect, the present invention provides a computer-readable storage medium, comprising: when the program is executed by a processor, the steps of implementing the method for rapid detection of transient faults of power devices of a DC charging pile are implemented.

[0045] The beneficial effects of the present invention are as follows: by collecting charging information in real time and using charging pile fault identification models and anomaly detection models, the charging characteristics of DC charging piles are comprehensively and deeply analyzed, helping operators to quickly detect fault types during the charging process, optimize the operation and management of charging piles, and improve operational efficiency and service quality. At the same time, by analyzing the data during the charging process, the layout of charging piles can be optimized, the user charging experience can be improved, and the operating costs can be reduced, providing a scientific and effective decision-making reference for the planning and policy formulation of future electric vehicle charging networks. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor. Among them:

[0047] Figure 1 A schematic diagram of the basic process of a method for rapid detection of transient faults of power devices in a DC charging pile provided by an embodiment of the present invention;

[0048] Figure 2 A technical circuit diagram of a method for rapid detection of transient faults of power devices in a DC charging pile provided by an embodiment of the present invention;

[0049] Figure 3A charging power difference curve diagram of a method for rapid detection of transient faults of power devices in a DC charging pile provided by an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of a charging pile fault identification model based on soft classification for a method for rapid detection of transient faults of power devices in a DC charging pile provided by an embodiment of the present invention;

[0051] Figure 5 A schematic diagram of a DC charging pile fault identification model for a method for rapid detection of transient faults in a DC charging pile power device provided by an embodiment of the present invention;

[0052] Figure 6 A schematic diagram of normal probability distribution and cumulative distribution of normal samples of a method for rapid detection of transient faults of power devices of a DC charging pile provided by an embodiment of the present invention;

[0053] Figure 7 A schematic diagram of the cumulative distribution of failure probabilities of normal samples of a method for rapid detection of transient failures of power devices in a DC charging pile provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0055] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0057] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0058] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0059] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0060] Example 1

[0061] Reference Figure 1 , is an embodiment of the present invention, and provides a method for rapid detection of transient faults of power devices of a DC charging pile, comprising:

[0062] S1: Collect the first feature of the DC charging pile;

[0063] S2: Processing and analyzing the collected first feature through the charging pile fault identification model, detecting the changes in each charging stage and analyzing the change rules;

[0064] S3: Based on the change rules, monitor the abnormal fluctuations of charging current and charging voltage during charging and classify the fault types;

[0065] S4: Analyze the abnormal fluctuation through an abnormality detection model to determine the fault type in the charging process.

[0066] In the embodiment of the present application, by real-time monitoring of the current and voltage of the DC charging pile, and applying data mining and pattern recognition technology, an unsupervised learning method is used to quickly detect possible power device failures or abnormal conditions, so as to intelligently sense failures during the charging process. At the same time, considering the battery health status and battery thermal coupling model, an optimized machine learning algorithm is used to analyze the charging pile failure data and battery failure mode, and evaluate the accuracy and reliability of various early warning methods.

[0067] Example 2

[0068] Reference Figure 2-5, which is an embodiment of the present invention, provides a method for rapid detection of transient faults of power devices of a DC charging pile based on the previous embodiment, comprising:

[0069] In the embodiments of the present application, the charging facility is a high-power output device, and its power devices may get out of control under conditions such as aging, causing safety hazards. In order to improve charging safety, transient fault prediction and rapid perception based on big data methods are crucial. By real-time monitoring of the current and voltage of the DC charging pile, and applying data mining and pattern recognition techniques, an unsupervised learning method is used to quickly detect possible power device failures or abnormal conditions, so as to intelligently sense faults during the charging process. At the same time, considering the battery health status and battery thermal coupling model, an optimized machine learning algorithm is used to analyze the charging pile fault data and battery failure mode, and evaluate the accuracy and reliability of various early warning methods. The technical circuit diagram for rapid detection of transient faults of DC charging pile power devices for charging process early warning is shown below. Figure 2 shown.

[0070] In the embodiments of the present application, Figure 2 As shown, intelligent sensing of charging pile faults is performed as follows:

[0071] According to the structure of the charging pile, the fault of the charging pile can be divided into multiple categories, such as AC side fault, converter fault, DC network fault, battery fault, etc. Specifically, the five types of faults represent DC bus fault, abnormality between charging poles, abnormality of converter DC side outlet, load side fault, and power supply side fault. Under normal circumstances, the probability of failure of the cable line of the charging pile is low, while the DC bus is more prone to failure. In addition, the failure of the DC converter may be cut off in time by the protection action of the internal switching element. As for the battery pack failure, it mainly includes short circuit, overload, aging and breakdown of the battery.

[0072] 1) Analysis of abnormal charging curve

[0073] Through the real-time monitoring of the charging current and voltage information at the charging pile end, the real-time charging power can be calculated to reflect the power output of the charging pile per unit time. At the same time, combined with the real-time SOC information from the vehicle end, the power input of the electric vehicle can be understood. During the normal charging process, the charging of the DC charging pile can be divided into three different stages, each of which has specific charging characteristics: Stage 1 corresponds to the vehicle-side SOC value in the range of 0% to 90%, at which time the charging power at the pile end remains at a relatively high level and increases slowly; Stage 2 corresponds to the vehicle-side SOC value generally in the range of 90% to 95%, at which time the charging power at the pile end will drop rapidly from the highest value to a lower value; Stage 3 corresponds to the vehicle-side SOC value close to 100%, at which time the charging power at the pile end will remain at the lowest level. In order to compare with the normal charging rules, Figure 3The charging power difference curves of some fault types are shown. During the fault charging process, the power difference value will drop significantly in advance and fluctuate greatly around 0. Such a change trend provides a powerful clue for identifying and monitoring the fault type of the charging pile.

[0074] 2) Construction of charging pile fault identification model

[0075] Reference Figure 4 The schematic diagram of the principle of the charging pile fault identification model based on soft classification is shown. With the help of the integration method of random forest (RF), K nearest neighbor (KNN) and extreme gradient boosting (XGBoost), a DC charging pile fault identification model is established. In this model, RF, KNN and XGBoost act as sub-classifiers, and each sub-classifier calculates the probability of each type of fault occurring for the input sample. The integrated model performs a weighted summation of the probabilities of each type of fault occurring in the samples calculated by each classifier, and thus determines the fault type with the highest summation result as the discrimination result of the identification model. This integration method combines the advantages of different sub-classifiers and improves the accuracy and robustness of the fault identification model.

[0076] During the data acquisition stage, the DC charging pile only outputs the current and voltage values ​​of phase A, while the current and voltage values ​​of phases B and C are always zero. The phase A current, phase A voltage, and SOC recording data of the DC charging pile during each charging process are regarded as independent sample information. For each sample, the maximum value in the current and voltage sequences is extracted, and the differential sequence of power is calculated, and then these values ​​are combined with the SOC sequence of the corresponding charging process to form the charging characteristics of the sample. At the same time, basic information such as the rated power and maximum output voltage of the DC charging pile is used as input features of the fault identification model. In order to establish a fault identification model, a kernel based on the "RF+KNN+XGBoost" integration is selected. The model is able to output the probability value of each type of fault. Overall, the framework diagram of the DC charging pile fault identification model is shown in the figure below. Figure 5 As shown in the figure, key information such as current, voltage, SOC, etc. are fully considered, providing a reliable basis for accurate fault identification.

[0077] In the embodiments of the present application, Figure 2 As shown in the figure, the charging warning research and evaluation indicators are studied, as follows:

[0078] The warning threshold is closely related to the accuracy and lead time of the warning: as the warning threshold increases, the warning accuracy of the model improves, and the credibility of the warning increases accordingly. However, as the warning threshold increases, the warning lead time shortens, and the actual advance notice time decreases, thereby reducing the actual value of the warning. Because different types of faults have their own characteristics, independent warning thresholds should be set for different fault types. In order to set these thresholds, statistical methods are used. By fitting the probability values ​​of various faults calculated by the model of normal samples of fault-free charging piles, the probability density function is obtained, and finally the corresponding warning threshold is calculated.

[0079] In the embodiments of the present application, Figure 2 As shown in the figure, in the charging current and voltage fault rapid detection and intelligent sensing module, by analyzing the charging characteristics, possible power device faults or abnormal conditions are inferred, and the charging current and voltage fault rapid detection and intelligent sensing are realized. Real-time monitoring of the AC and DC information of the DC charging pile is as follows:

[0080] 1) Analysis of power device failure or abnormal status

[0081] As a key power facility for electric vehicles, charging piles have a harsh working environment and may encounter a variety of power device failures or abnormal conditions during daily operation. These situations may include:

[0082] (a) Overload. Long-term high-load operation, such as high-power charging demand exceeding the design load capacity of the charging pile, may cause thermal damage to circuit boards, overload of electronic components, and even damage to key components.

[0083] (b) Overheating. Prolonged operation in a high temperature environment or poor heat dissipation may cause internal components to overheat, resulting in electronic component failure or even damage to cables or connectors.

[0084] (c) Circuit short circuit or open circuit. A short circuit or open circuit in the internal circuit components of the charging pile may cause the charging pile to operate abnormally, such as charging interruption or reduced charging rate.

[0085] (d) Abnormal voltage. Abnormal voltage fluctuations may indicate internal circuit problems, such as a power management unit failure, which may affect charging efficiency or safety.

[0086] (e) Abnormal current. Abnormal current output may indicate a current sensor failure or power supply problem, which may affect charging efficiency or safety.

[0087] (f) Communication failure. Failure in communication between the charging pile and the external system may result in data transmission errors or the inability to remotely monitor and manage the charging pile.

[0088] (g) Controller failure. A failure in the controller unit or related circuits may result in inefficient charging or failure to complete charging normally.

[0089] According to the structure of the charging pile, the fault of the charging pile is divided into AC side fault, converter fault, DC network fault, battery fault, etc. As shown in the figure, the five faults represent DC bus fault, abnormal charging pole, abnormal converter DC side outlet, load side fault, and power supply side fault. Generally speaking, the probability of failure of the cable line of the charging pile is low; while the DC bus is more prone to failure; the failure of the DC converter can be timely removed by the protection action of the internal switching element; the battery pack failure is mainly short circuit, overload, aging and breakdown of the battery.

[0090] 2) Current and voltage monitoring and charging characteristics analysis

[0091] Real-time data acquisition, using sensors to monitor the current and voltage of DC charging piles in real time. Current sensors and voltage sensors are embedded inside the charging pile or on the charging line to capture charging AC and DC information and convert analog signals into digital signals.

[0092] Data analysis and feature extraction: Process and analyze the collected current and voltage data. Use data processing techniques, including filtering and Fourier transform, to extract charging features, such as charging stage, stability, etc. Detect changes in charging stages, such as the startup stage, constant current charging stage, and charging completion stage, and analyze their changing patterns.

[0093] Anomaly detection and fault diagnosis, establish anomaly detection models, monitor abnormal fluctuations in current and voltage during charging or patterns that do not conform to expectations. By comparing with normal patterns, possible power device failures or abnormal conditions can be identified. Machine learning algorithms such as support vector machines and neural networks are used to quickly detect and diagnose abnormalities in charging current and voltage.

[0094] Intelligent sensing and fast detection, using intelligent sensing technology, building models to sense and analyze charging current and voltage in real time. Using fast detection algorithms, such as threshold-based anomaly detection, to achieve fast detection of current and voltage anomalies.

[0095] The above method, combined with real-time data collection and analysis, anomaly detection and the establishment of fault diagnosis models, can monitor the charging AC and DC information of DC charging piles and analyze the charging characteristics. At the same time, it can infer possible power device failures or abnormal conditions and realize rapid detection and intelligent sensing of charging current and voltage.

[0096] 3) Application of unsupervised learning algorithms

[0097] Unsupervised learning is a branch of machine learning that aims to discover patterns and structures from unlabeled data. Its main principle is based on the statistical properties of data, without relying on labels or known results. Based on this, unsupervised learning can be used to infer power device failures or abnormal conditions and label them. The operation process includes:

[0098] (a) Data collection and preprocessing. First, data is collected and preprocessed. This involves cleaning the data, filling in missing values, dealing with outliers, and performing operations such as standardization and normalization to ensure data quality and consistency for algorithm processing.

[0099] (b) Select algorithms and models. Next, select the appropriate unsupervised learning algorithm or model based on the task and data characteristics. For example, for clustering tasks, you can choose the K-means clustering algorithm; for dimensionality reduction tasks, you can use algorithms such as principal component analysis (PCA).

[0100] (c) Model training. The selected algorithm is applied to the data for training. The algorithm attempts to find patterns or structures in the data, fits them according to the characteristics of the data, and creates a corresponding model.

[0101] (d) Evaluation and analysis. Evaluate the performance of the model and interpret the generated patterns or clusters to obtain the underlying information behind the data. Evaluation can include internal indicators (such as silhouette coefficient) and external indicators (such as clustering accuracy), while interpretation involves analysis and understanding of clustering results or feature dimensionality reduction.

[0102] Unsupervised learning has a wide range of applications in many fields. Its cluster analysis can group data into similar groups, which is suitable for market segmentation and social network analysis. Anomaly detection can find abnormal patterns in data, such as network security and financial fraud detection. At the same time, it performs equally well in dimensionality reduction and feature extraction, and is used for data compression and extracting the most relevant features, such as image and speech recognition. Through association rule mining, it helps to discover correlations in data, such as shopping cart analysis and recommendation systems. Unsupervised learning is a powerful tool for processing large amounts of unlabeled data and discovering potential information and knowledge, providing a powerful means for in-depth understanding of data structures and patterns.

[0103] 4) Inference of power device failure or abnormal state

[0104] 4.1 Methods for diagnosing fault or abnormal state of charging pile power device

[0105] Real-time monitoring and data analysis. Use sensors to monitor the current, voltage and other data of charging piles in real time. Through data analysis techniques such as anomaly detection and pattern recognition, irregular patterns or abnormal situations can be found. Use time series analysis, machine learning algorithms, etc. to detect anomalies and identify possible fault signals.

[0106] Fault codes and alarms. Charging stations are usually designed with fault codes or alarm systems. When the system detects an abnormal situation, a corresponding code or alarm will be generated to remind you that there is a problem and to conduct further inspection or processing.

[0107] Remote diagnosis. Through the remote monitoring system, the data and status of the charging pile can be accessed remotely. This method allows remote diagnosis and abnormal analysis, timely detection and resolution of potential problems, and reduces the impact of failures on users.

[0108] Professional testing and maintenance. Regular professional equipment testing and maintenance is the key to preventing failures. Regular equipment inspections, maintenance and upkeep, and regular testing and troubleshooting by professional maintenance personnel can help prevent and resolve potential problems and improve the reliability and continued operation of charging facilities.

[0109] By combining the above methods, faults or abnormal conditions of charging pile power devices can be discovered and diagnosed in a timely manner, ensuring safe and stable operation of the equipment.

[0110] 4.2 Based on the unsupervised learning algorithm, the power device fault or abnormal state is inferred to achieve rapid detection and intelligent sensing of charging current and voltage. The inference process based on the unsupervised learning algorithm is as follows:

[0111] a) Data collection and preprocessing: Collect real-time charging current and voltage data and perform preprocessing, including removing outliers and standardizing data, to ensure data quality.

[0112] b) Feature extraction. Useful features are extracted from current and voltage data. Frequency domain features (such as frequency, spectrum), time domain features (such as waveform, periodicity), etc. can be used to describe the characteristics of current and voltage.

[0113] c) Anomaly detection model construction: Use unsupervised learning algorithms such as density-based anomaly detection (such as DBSCAN), distance-based anomaly detection (such as LOF, KNN), or clustering-based methods (such as K-means) to model and train the current and voltage data.

[0114] d) Anomaly detection and intelligent sensing. During real-time monitoring, the current and voltage data collected in real time are input into the constructed anomaly detection model. The model will identify data points that do not conform to the normal pattern and mark them as anomalies. These anomalies may indicate potential problems, such as equipment failure, circuit problems, etc.

[0115] e) Rapid response and processing. Once an abnormality is detected, the system can automatically trigger an alarm or notify relevant personnel so that they can respond quickly and take appropriate measures. For example, interrupt charging, safely stop equipment operation, notify maintenance personnel, etc.

[0116] f) Model optimization and feedback: Based on actual operation conditions and anomaly feedback, the anomaly detection model is continuously optimized and updated to improve the accuracy and reliability of detection.

[0117] The anomaly detection model built using an unsupervised learning algorithm can quickly detect and intelligently sense abnormal conditions of charging current and voltage, providing support for real-time monitoring and fault diagnosis, thereby ensuring the safe operation of charging equipment.

[0118] In the embodiments of the present application, Figure 2 As shown in the figure, in the charging warning rule formulation and optimization module, by formulating and optimizing the charging warning rules, the accuracy and reliability of different warning methods are evaluated. According to the charging pile and battery thermal coupling model, the battery charging SOC change is considered. The three steps are battery safety and status monitoring, battery charging physical model construction, and charging warning rule formulation and optimization. The details are as follows:

[0119] 1) Research on thermal coupling model of charging pile and battery

[0120] In the development of charging pile and battery technology, the establishment of thermal coupling model becomes the key to achieve effective charging management and battery protection. This model involves many aspects, from the measurement of thermophysical parameters to the establishment of heat conduction model, to the development and optimization of thermal coupling algorithm. The following will study the key steps of this process and introduce the key considerations for modeling in this field.

[0121] (a) Thermophysical parameter measurement. First, parameter acquisition requires obtaining the thermophysical parameters of the charging pile and battery, such as thermal conductivity, specific heat capacity, etc. These parameters can be obtained through laboratory tests or literature research. Then, experimental verification is performed to verify the parameters used in the model to ensure their accuracy and reliability. For example, thermal conductivity can be obtained through thermal container experiments or heat conduction experiments.

[0122] (b) Temperature sensor installation. First, sensor location selection and layout. Install temperature sensors on charging piles and batteries, and select appropriate locations to accurately monitor temperature changes. Optimizing sensor layout can capture changes in temperature distribution to the greatest extent. Then, accurate data collection. The collected temperature data needs to have high accuracy and high sampling rate in order to establish an accurate thermal coupling model.

[0123] (c) Establishment of heat conduction model. First, model selection. The heat conduction model can be established by using heat conduction equation or finite element analysis, taking into account the heat exchange process and thermal coupling effect between the charging pile and the battery. Then, parameter adjustment. By comparing the model with actual data, the model parameters are adjusted to improve its accuracy and fidelity.

[0124] (d) Thermal coupling algorithm. First, the model is integrated to develop a thermal coupling algorithm by combining the thermal physics models of the battery and the charging pile to simulate the temperature changes during the charging process. Then, the thermal management strategy is developed to use the algorithm for thermal management control to optimize charging efficiency, extend battery life, and predict and solve potential thermal problems.

[0125] (e) Verification and optimization. First, experimental data verification is performed to compare the established thermal coupling model with actual data, and the parameters and assumptions in the model are corrected to improve the accuracy and reliability of the model. Then, iterative optimization is performed to continuously optimize the model, using actual feedback and continuously collected data to ensure that the model is consistent with the actual situation and is predictive.

[0126] 2) Battery SOH estimation and SOC estimation

[0127] (a) Estimation of the state of health (SOH) of lithium-ion batteries is an important indicator for evaluating battery life and performance degradation. Considering the battery health state and integrating the battery charging SOC changes during charging can reduce charging pile failures and battery failures. The following methods are usually used to estimate SOH:

[0128] a) Capacity decay analysis. Through periodic charge and discharge tests, the number of charge and discharge cycles and capacity loss of the battery are recorded. By tracking the difference between the rated capacity and the actual capacity of the battery, the capacity decay rate can be estimated, and thus the SOH can be inferred.

[0129] b) Internal resistance test. Internal resistance has the greatest impact on battery performance. Internal resistance test is a commonly used evaluation method. It can measure internal resistance through AC impedance spectroscopy and other technologies, and evaluate the health status of the battery accordingly.

[0130] c) Temperature and cycle monitoring. Track the operating temperature and cycle number of the battery. High temperature and excessive charge and discharge cycles will cause battery aging. Recording these parameters helps evaluate the health of the battery.

[0131] d) Prediction model. Based on physical models or data-driven methods, historical data and monitoring parameters are used to predict the health status of the battery. These models can include machine learning and statistical methods such as neural networks and regression analysis.

[0132] e) Comprehensive evaluation: Combine multiple monitoring parameters and models and use comprehensive evaluation methods, such as weighted average or multi-indicator fusion, to evaluate the overall health status of the battery.

[0133] (b) Estimation of the state of charge (SOC) of lithium-ion batteries is an important part of implementing a battery management system (BMS). The battery SOC is estimated in real time during the charging process, and the charging strategy is adjusted in time to better ensure battery safety and extend battery life. The following are some common methods for estimating SOC:

[0134] a) Voltage method. SOC is estimated by voltage measurement. This is one of the most commonly used methods, which uses the relationship between battery voltage and SOC during charging and discharging. However, due to factors such as charge and discharge rate, temperature and capacity attenuation, the estimation accuracy needs to be improved.

[0135] b) Ampere-hour method. Estimate SOC based on the accumulation of battery charge and discharge. This method uses the product of current and time to track the charge and discharge of the battery and estimates SOC through calculation. As time increases, the error of this method may accumulate.

[0136] c) Filtering and Kalman filtering. These methods use filters to process the measured data such as current and voltage to improve the estimation accuracy of SOC. Kalman filtering is particularly suitable for dynamic systems and can take into account prior information and noise to improve estimation accuracy.

[0137] d) Equivalent circuit model method. Based on the equivalent circuit model of the battery, combined with current and voltage measurement data, the SOC is estimated using parameters such as battery internal resistance, battery voltage, and open circuit voltage. This method is usually more accurate, but requires appropriate model parameters.

[0138] e) Deep learning and machine learning based on comprehensive methods. Deep learning and machine learning models are used to combine multiple data sources and sensor measurements to predict and estimate SOC. These methods can use large amounts of data to improve accuracy and consider more complex battery behaviors.

[0139] 3) Construction of battery charging physical model

[0140] Building a battery charging physical model based on machine learning algorithms is an important research direction, which aims to improve the safety of the battery charging process and warn of potential problems. By collecting, converting and analyzing a large amount of charging data, a reliable physical model is built, and its predictive ability is used to provide early warning and control for the battery charging process, thereby improving the safety and efficiency of battery charging.

[0141] (a) Data collection. Collect data during the battery charging process, including charging voltage, current, temperature and other parameters. The data should cover different battery types, working conditions and charging modes to ensure that the model is representative and generalizable.

[0142] (b) Feature extraction and selection. Based on the collected data, feature extraction and selection are performed to convert the raw data into a feature set acceptable to the model. This includes frequency domain or time domain feature extraction, as well as related feature selection methods.

[0143] (c) Model selection and training. Select appropriate data-driven methods, such as neural networks, support vector machines, regression models, etc., to build the model. Use the prepared dataset to train the model.

[0144] (d) Model evaluation and optimization. Evaluate the model and evaluate its performance using cross-validation or a holdout validation set, such as k-fold cross-validation. Optimize the model based on its predictive accuracy, generalization ability, and stability. You may need to adjust model parameters or try different model architectures.

[0145] (e) Model verification and application. Finally, the optimized model is verified to test its performance on new data. If the model meets expectations and has good predictive ability, it can be deployed in practical applications for physical model construction and prediction of the battery charging process.

[0146] 4) Formulation and optimization of charging warning rules

[0147] (a) Charging-related data collection and analysis

[0148] The safety and performance of charging piles cannot be separated from compliance standards and regulations. When formulating charging warning rules, it is crucial to understand and comply with relevant charging standards and regulations. Data collection and analysis are the basis, which involves collecting real-time parameters of charging piles and batteries, and at the same time, through historical data analysis, mastering the typical behavior patterns and abnormal characteristics of batteries under different charging states. In addition, considering the impact of charging standards (regulations), sufficient charging standard research and compliance requirements analysis are carried out to ensure that the charging process meets safety and regulatory requirements. Comprehensive consideration of standards and data analysis will help to formulate more reliable and safe charging warning rules to ensure the safety and compliance of the charging process.

[0149] (b) Charging warning rule formulation

[0150] When determining the warning rules, it is necessary to consider the warning rules under different battery states. Indicators such as SOC (state of charge) and SOH (state of health) can be used to set state assessment rules. For example, a warning is triggered when the SOC is below a certain threshold or the SOH drops sharply. In addition, it is also crucial to formulate rules for monitoring abnormal values, covering abnormal temperatures, currents, voltages, etc. to identify abnormal charging states. In addition, safety factors should be considered when setting charging process rules. There may be potential risks in situations such as charging rates that are too fast, excessive temperatures generated during charging, and frequent disconnection of charging piles. Corresponding rules should be formulated to warn and avoid the occurrence of potentially dangerous situations. The setting of these rules needs to take into account the actual conditions in different scenarios, and make reasonable adjustments based on experience and expertise to ensure the safety, stability and compliance of the charging process.

[0151] (c) Optimization of charging warning rules

[0152] By using intelligent search algorithms such as genetic algorithms and rule engines combined with expert knowledge and empirical rules, charging warning rules can be optimized and adjusted. First, a large amount of data generated during the charging process is collected and used as input for the optimization algorithm. Through intelligent algorithms such as genetic algorithms, optimization models can be established based on a large amount of data and expert knowledge, and then these models can be used to evaluate and adjust existing warning rules.

[0153] Genetic algorithms can be used to search for optimal solutions. When adjusting warning rules, they can simulate the process of biological evolution, generate and modify the parameters of warning rules, and optimize the parameters and thresholds of rules to improve the performance and accuracy of rules. Rule engines can help manage and execute these rules.

[0154] Expert knowledge and empirical rules refer to rules set based on experts’ understanding of the charging process and battery behavior. These rules can be empirical, based on industry standards, or obtained from actual production environments. Combining these rules with intelligent algorithms can better improve the flexibility and adaptability of warning rules, making them better suited to different charging scenarios and situations.

[0155] 5) Research on early warning of charging pile failure and battery failure mode

[0156] In the context of the rapid development of electric vehicle technology today, charging pile failure and battery failure have a significant impact on the performance of electric vehicles. Analyzing charging pile failure data and battery failure modes, and evaluating the accuracy and reliability of different early warning methods have become crucial tasks. In this process, several key steps are involved:

[0157] (a) Data collection and cleaning. Collect historical data of charging piles and batteries, including parameters such as current, voltage, temperature, and charging rate during the charging process, as well as records of the operating status of the charging piles. When cleaning the data, pay attention to handling outliers and missing data to ensure data quality.

[0158] (b) Failure mode analysis. Use data analysis techniques, such as statistical methods, cluster analysis, and anomaly detection algorithms, to explore the failure modes in historical data. By mining data, different types of failure modes and possible failure characteristics can be identified to gain a deeper understanding of possible problems.

[0159] (c) Early warning method evaluation. Select appropriate early warning methods to respond to the identified failure modes. This may involve the use of rule engines, machine learning models (such as decision trees, neural networks), anomaly detection algorithms, etc. Evaluate the performance of different methods and examine their accuracy and practicality in fault detection and prediction.

[0160] (d) Model verification and validation. Apply the selected early warning method to real data and compare its prediction results with the actual failure conditions. This verification helps to evaluate the accuracy and reliability of the prediction model, discover the limitations of the model and make corrections.

[0161] (e) Improvement and optimization. Based on the verification results, the early warning model is adjusted and optimized. It may be necessary to adjust model parameters, introduce new features, improve data processing methods, etc. to improve prediction accuracy and fault detection sensitivity.

[0162] (f) Continuous monitoring and updating. As the usage of charging piles and batteries changes, the performance of the early warning method is continuously monitored. The model and data set are updated regularly, and the early warning system is continuously improved to adapt to changes in actual conditions.

[0163] This embodiment also provides a DC charging pile power device transient fault rapid detection system, including:

[0164] A collection module, used for collecting the first feature of the DC charging pile;

[0165] A detection module, used to process and analyze the collected first feature through a charging pile fault identification model, detect changes in each charging stage and analyze the change rules;

[0166] A classification module is used to monitor abnormal fluctuations of charging current and charging voltage during charging and classify fault types based on the change rules;

[0167] The analysis module is used to analyze the abnormal fluctuation through an abnormal detection model to determine the fault type in the charging process.

[0168] Furthermore, it also includes:

[0169] Memory, used to store programs;

[0170] A processor is used to load the program to execute the method for rapid detection of transient faults of power devices of a DC charging pile.

[0171] This embodiment also provides a computer-readable storage medium storing a program, which, when executed by a processor, implements the method for rapid detection of transient faults of power devices of a DC charging pile.

[0172] The storage medium proposed in this embodiment and the method for rapid detection of transient faults of DC charging pile power devices proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be referred to the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0173] Through the above description of the implementation methods, the technicians in the relevant field can clearly understand that the present invention can be implemented by means of software and necessary general hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ReadOnly, Memory, ROM), random access memory (RandomAccess Memory, RAM), flash memory (FLASH), hard disk or optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform the methods of various embodiments of the present invention.

[0174] Example 3

[0175] Reference Figure 6-7 Tables 1 and 2 are an embodiment of the present invention, which provides a method for rapid detection of transient faults of power devices in a DC charging pile. In order to verify its beneficial effects, comparison results of multiple solutions are provided.

[0176] In order to evaluate the performance of the DC charging pile fault identification model, accuracy (Precision), recall (Recall) and F1-Score are used as evaluation indicators. The larger the values ​​of these three indicators, the better the model performance. Considering that the model involves the identification of multiple fault labels, the Macro Average (numerical average) and MicroAverage (weighted average) values ​​of each evaluation indicator are calculated as a comprehensive evaluation of the performance of the fault identification model. The numerical average reflects the average performance of various types of faults on the same type of indicators, while the weighted average comprehensively considers the proportion information of various types of samples, thereby comprehensively reflecting the performance of the model on the entire data set.

[0177] The following Table 1 summarizes the number of 12 types of charging pile faults. The data includes historical information of more than 6,000 DC charging piles, which covers real-time collection, fault reports, and equipment information, including real-time voltage, real-time current, and real-time SOC data. The real-time power is obtained by calculation based on real-time current and voltage, and then the power is processed by first-order difference to analyze the charging power of the DC charging pile. According to the statistical results, the data involves more than 40 types of faults. After selection by experienced staff, the fault types with higher fault frequency and relative importance are focused on, which helps to gain a deeper understanding of the types of charging pile faults and their frequency of occurrence.

[0178] Table 1 Fault type statistics

[0179]

[0180]

[0181] The maximum current, maximum voltage, power differential sequence, SOC sequence, rated power and maximum allowable output voltage of each charging process of the DC charging pile are used as the input features of a sample. The samples located in the start and end time periods of the fault are marked by the fault information. During the fault time period, if there is no relevant charging record, that is, the charging pile stops providing charging service due to a fault, the charging process sample closest to the start time of the fault is marked as a fault sample. The training set and the test set are divided into a ratio of 7:3, and the five-fold cross-validation method is used for training. The model parameters are optimized through grid parameter adjustment. In the case of the same data set, the integrated model proposed in this embodiment is compared with KNN, XGBoost, and RF. The training and testing results of each model are shown in Table 2 below:

[0182] Table 2 Model performance comparison

[0183]

[0184]

[0185] The results show that compared with the fault identification model of a single algorithm, the integrated model has higher accuracy and better performance, which verifies the effectiveness of the model in charging pile fault diagnosis.

[0186] In addition, the integrated model is used to predict the charging pile samples that have never failed. The model presents the probability distribution and probability cumulative distribution of the above normal samples that no failure will occur, as shown in the following figure. Figure 6 The normal probability distribution and cumulative distribution diagram of normal samples are shown in Figure 2. In this case, 90% of the normal samples are predicted by the model to have a probability value of more than 0.8 without fault occurrence, which shows that the model has a high accuracy in identification.

[0187] In addition, in the prediction of the model, the cumulative distribution of the probability of various types of failures occurring in the above normal samples is shown in Figure 7 In the diagram of the cumulative distribution of the failure probability of normal samples, it is observed that the cumulative distribution of various types of failures is close to 100% at locations where the model predicts a lower probability value, which further confirms the accuracy of the model.

[0188] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for rapid detection of transient faults of power devices in a DC charging pile, characterized in that: include: Collect the first feature of DC charging pile; The collected first feature is processed and analyzed by the charging pile fault identification model to detect the changes in each charging stage and analyze the change rules; Based on the changing rules, monitor the abnormal fluctuations of charging current and charging voltage during charging and classify the fault types; The abnormal fluctuation is analyzed by an abnormal detection model to determine the type of fault in the charging process.

2. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 1, characterized in that: The first characteristics include the charging current and charging voltage of the DC charging pile, and the charge state information of the charging vehicle.

3. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 1 or 2, characterized in that: The charging pile fault identification model comprises the following steps: Random Forest, K-Nearest Neighbors, and Extreme Gradient Boosting were used as sub-classifiers; Each sub-classifier calculates the probability of each type of fault occurring for the first feature of the input; By taking a weighted sum of the probabilities of each type of fault occurring in the first feature calculated by each classifier, the fault type with the highest summation result is determined as the discrimination result of the recognition model.

4. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 3, characterized in that: The charging stage includes a start-up stage, a constant current charging stage and a charging completion stage; The classification of fault types includes DC bus fault, charging pole abnormality, converter DC side outlet abnormality, load side fault and power supply side fault.

5. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 4, characterized in that: The detecting the changes in each charging stage and analyzing the changing rules thereof comprises the following steps: If the charging power fluctuates abnormally during the startup phase, it indicates a DC bus fault; If the charging power fluctuates abnormally during the startup phase, it indicates a power supply side fault; If the charging power growth stagnates during the startup phase, it indicates a load-side fault; If the charging power drops rapidly and sharply during the constant current charging phase and is lower than the preset safety threshold, it indicates that there is an abnormality at the DC side outlet of the converter; If the charging power fluctuates frequently and cannot drop to the expected minimum point during the charging completion stage, it indicates that there is an abnormality between the charging poles.

6. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 5, characterized in that: The anomaly detection model includes using an unsupervised learning algorithm to model and train the current and voltage data.

7. The method for rapid detection of transient faults of power devices of a DC charging pile according to claim 6, characterized in that: The analyzing the abnormal fluctuation by using an abnormality detection model includes inputting the real-time collected current and voltage data into the constructed abnormality detection model, identifying data points that do not conform to the normal pattern, and marking them as abnormal.

8. A system based on the method for rapid detection of transient faults of power devices of a DC charging pile according to claim 1, characterized in that: A collection module, used for collecting the first feature of the DC charging pile; A detection module, used to process and analyze the collected first feature through a charging pile fault identification model, detect changes in each charging stage and analyze the change rules; A classification module is used to monitor abnormal fluctuations of charging current and charging voltage during charging and classify fault types based on the change rules; The analysis module is used to analyze the abnormal fluctuation through an abnormal detection model to determine the fault type in the charging process.

9. An electronic device, characterized in that: include: Memory, used to store programs; A processor is used to load the program to execute the steps of the method for rapid detection of transient faults of a power device of a DC charging pile as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a program, characterized in that: When the program is executed by the processor, the steps of the method for rapid detection of transient faults of a DC charging pile power device as described in any one of claims 1 to 7 are implemented.

Images